Neurocomputing-neurons as microcomputers

Studies on neurocomputing should be directed in two ways which in turn influence each other. In one direction concrete neural-network solutions for specific important problems should be applied to substantiate its practical significance and, at the same time, the theoretical potentialities and limitations for computation be explored on such network models. The alternative way is to seek a more profound understanding of the algorithms used by the CNS (the central nervous system) to process informations and to know more about the molecular and cellular mechanisms underlying specific computations and memory process in neurons and neural networks. From this point of view, artificial neurons developed for neurocomputing are oversimplified to simulate real neurons. Here I will show that neurons are analogous to microcomputers whose characteristics can be classified into over 50 kinds, and that electrical events observed in neurons are a part of the many manifestations associated with neural activities and are regulated by the chemical and conformational processes inside neurons.